The ctx structure passed into bpf programs is different depending on bpf
program type. The verifier incorrectly marked ctx->data and ctx->data_end
access based on ctx offset only. That caused loads in tracing programs
int bpf_prog(struct pt_regs *ctx) { .. ctx->ax .. }
to be incorrectly marked as PTR_TO_PACKET which later caused verifier
to reject the program that was actually valid in tracing context.
Fix this by doing program type specific matching of ctx offsets.

Fixes: 969bf05e ("bpf: direct packet access")
Reported-by: NSasha Goldshtein <goldshtn@gmail.com>
Signed-off-by: NAlexei Starovoitov <ast@kernel.org>
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

1) gre_parse_header() can be called from gre_err()

   At this point transport header points to ICMP header, not the inner
header.

2) We can not really change transport header as ipgre_err() will later
assume transport header still points to ICMP header (using icmp_hdr())

3) pskb_may_pull() logic in gre_parse_header() really works
  if we are interested at zone pointed by skb->data

4) As Jiri explained in commit b7f8fe25 ("gre: do not pull header in
ICMP error processing") we should not pull headers in error handler.

So this fix :

A) changes gre_parse_header() to use skb->data instead of
skb_transport_header()

B) Adds a nhs parameter to gre_parse_header() so that we can skip the
not pulled IP header from error path.
  This offset is 0 for normal receive path.

C) remove obsolete IPV6 includes
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Tom Herbert <tom@herbertland.com>
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Jiri Benc <jbenc@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The implementation of net_dbg_ratelimited in the CONFIG_DYNAMIC_DEBUG
case was added with 2c94b537 ("net: Implement net_dbg_ratelimited() for
CONFIG_DYNAMIC_DEBUG case"). The implementation strategy was to take the
usual definition of the dynamic_pr_debug macro, but alter it by adding a
call to "net_ratelimit()" in the if statement. This is, in fact, the
correct approach.

However, while doing this, the author of the commit forgot to surround
fmt by pr_fmt, resulting in unprefixed log messages appearing in the
console. So, this commit adds back the pr_fmt(fmt) invocation, making
net_dbg_ratelimited properly consistent across DEBUG, no DEBUG, and
DYNAMIC_DEBUG cases, and bringing parity with the behavior of
dynamic_pr_debug as well.

Fixes: 2c94b537 ("net: Implement net_dbg_ratelimited() for CONFIG_DYNAMIC_DEBUG case")
Signed-off-by: NJason A. Donenfeld <Jason@zx2c4.com>
Cc: Tim Bingham <tbingham@akamai.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The skfp driver has been moved to drivers/net/fddi/skfp a long time
ago, but we still attempt to include headers from the old location,
which causes a warning when building with W=1:

cc1: error: /git/arm-soc/drivers/net/skfp: No such file or directory [-Werror=missing-include-dirs]
cc1: error: drivers/net/skfp: No such file or directory [-Werror=missing-include-dirs]

Clearly this include directive is not needed any more, so we can
just remove it now.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

net/tipc/link.c: In function ‘tipc_link_timeout’:
net/tipc/link.c:744:28: warning: ‘mtyp’ may be used uninitialized in this function [-Wuninitialized]

Fixes: 42b18f60 ("tipc: refactor function tipc_link_timeout()")
Acked-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When run tipcTS&tipcTC test suite, the following complaint appears:

[   56.926168] ===============================
[   56.926169] [ INFO: suspicious RCU usage. ]
[   56.926171] 4.7.0-rc1+ #160 Not tainted
[   56.926173] -------------------------------
[   56.926174] net/tipc/bearer.c:408 suspicious rcu_dereference_protected() usage!
[   56.926175]
[   56.926175] other info that might help us debug this:
[   56.926175]
[   56.926177]
[   56.926177] rcu_scheduler_active = 1, debug_locks = 1
[   56.926179] 3 locks held by swapper/4/0:
[   56.926180]  #0:  (((&req->timer))){+.-...}, at: [<ffffffff810e79b5>] call_timer_fn+0x5/0x340
[   56.926203]  #1:  (&(&req->lock)->rlock){+.-...}, at: [<ffffffffa000c29b>] disc_timeout+0x1b/0xd0 [tipc]
[   56.926212]  #2:  (rcu_read_lock){......}, at: [<ffffffffa00055e0>] tipc_bearer_xmit_skb+0xb0/0x2e0 [tipc]
[   56.926218]
[   56.926218] stack backtrace:
[   56.926221] CPU: 4 PID: 0 Comm: swapper/4 Not tainted 4.7.0-rc1+ #160
[   56.926222] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2007
[   56.926224]  0000000000000000 ffff880016803d28 ffffffff813c4423 ffff8800154252c0
[   56.926227]  0000000000000001 ffff880016803d58 ffffffff810b7512 ffff8800124d8120
[   56.926230]  ffff880013f8a160 ffff8800132b5ccc ffff8800124d8120 ffff880016803d88
[   56.926234] Call Trace:
[   56.926235]  <IRQ>  [<ffffffff813c4423>] dump_stack+0x67/0x94
[   56.926250]  [<ffffffff810b7512>] lockdep_rcu_suspicious+0xe2/0x120
[   56.926256]  [<ffffffffa00051f1>] tipc_l2_send_msg+0x131/0x1c0 [tipc]
[   56.926261]  [<ffffffffa000567c>] tipc_bearer_xmit_skb+0x14c/0x2e0 [tipc]
[   56.926266]  [<ffffffffa00055e0>] ? tipc_bearer_xmit_skb+0xb0/0x2e0 [tipc]
[   56.926273]  [<ffffffffa000c280>] ? tipc_disc_init_msg+0x1f0/0x1f0 [tipc]
[   56.926278]  [<ffffffffa000c280>] ? tipc_disc_init_msg+0x1f0/0x1f0 [tipc]
[   56.926283]  [<ffffffffa000c2d6>] disc_timeout+0x56/0xd0 [tipc]
[   56.926288]  [<ffffffff810e7a68>] call_timer_fn+0xb8/0x340
[   56.926291]  [<ffffffff810e79b5>] ? call_timer_fn+0x5/0x340
[   56.926296]  [<ffffffffa000c280>] ? tipc_disc_init_msg+0x1f0/0x1f0 [tipc]
[   56.926300]  [<ffffffff810e8f4a>] run_timer_softirq+0x23a/0x390
[   56.926306]  [<ffffffff810f89ff>] ? clockevents_program_event+0x7f/0x130
[   56.926316]  [<ffffffff819727c3>] __do_softirq+0xc3/0x4a2
[   56.926323]  [<ffffffff8106ba5a>] irq_exit+0x8a/0xb0
[   56.926327]  [<ffffffff81972456>] smp_apic_timer_interrupt+0x46/0x60
[   56.926331]  [<ffffffff81970a49>] apic_timer_interrupt+0x89/0x90
[   56.926333]  <EOI>  [<ffffffff81027fda>] ? default_idle+0x2a/0x1a0
[   56.926340]  [<ffffffff81027fd8>] ? default_idle+0x28/0x1a0
[   56.926342]  [<ffffffff810289cf>] arch_cpu_idle+0xf/0x20
[   56.926345]  [<ffffffff810adf0f>] default_idle_call+0x2f/0x50
[   56.926347]  [<ffffffff810ae145>] cpu_startup_entry+0x215/0x3e0
[   56.926353]  [<ffffffff81040ad9>] start_secondary+0xf9/0x100

The warning appears as rtnl_dereference() is wrongly used in
tipc_l2_send_msg() under RCU read lock protection. Instead the proper
usage should be that rcu_dereference_rtnl() is called here.

Fixes: 5b7066c3 ("tipc: stricter filtering of packets in bearer layer")
Acked-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Sabrina Dubroca says:

====================
macsec fixes

Patch 1 adds rcu_barrier() during module unload to prevent possible
panics.

Patch 2 allocates memory for scattergather lists and the IV on the
heap, since they can escape the current function's context during
crypto callbacks.

Patch 3 fixes a failure to create secure associations.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The ASYNC flag prevents initialization on some physical machines.

Fixes: c09440f7 ("macsec: introduce IEEE 802.1AE driver")
Signed-off-by: NSabrina Dubroca <sd@queasysnail.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For the crypto callbacks to work properly, we cannot have sg and iv on
the stack.  Use kmalloc instead, with a single allocation for
aead_request + scatterlist + iv.

Fixes: c09440f7 ("macsec: introduce IEEE 802.1AE driver")
Signed-off-by: NSabrina Dubroca <sd@queasysnail.net>
Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Without this, the various uses of call_rcu could cause a kernel panic.

Fixes: c09440f7 ("macsec: introduce IEEE 802.1AE driver")
Signed-off-by: NSabrina Dubroca <sd@queasysnail.net>
Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

saw a debug splat:
net/include/net/sch_generic.h:287 suspicious rcu_dereference_check() usage!
other info that might help us debug this:
rcu_scheduler_active = 1, debug_locks = 0
 2 locks held by kworker/2:1/710:
  #0:  ("events"){.+.+.+}, at: [<ffffffff8106ca1d>]
  #1:  ((&q->work)){+.+...}, at: [<ffffffff8106ca1d>] process_one_work+0x14d/0x690
Workqueue: events htb_work_func
Call Trace:
 [<ffffffff812dc763>] dump_stack+0x85/0xc2
 [<ffffffff8109fee7>] lockdep_rcu_suspicious+0xe7/0x120
 [<ffffffff814ced47>] htb_work_func+0x67/0x70
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Acked-by: NCong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This refers to commands to direct action access as follows:

sudo tc actions add action drop index 12
sudo tc actions add action pipe index 10

And then dumping them like so:
sudo tc actions ls action gact

iproute2 worked because it depended on absence of TCA_ACT_TAB TLV
as end of message.
This fix has been tested with iproute2 and is backward compatible.
Signed-off-by: NJamal Hadi Salim <jhs@mojatatu.com>
Acked-by: NCong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

And avoid calling tcf_hash_check() twice.

Fixes: a57f19d3 ("net sched: ipt action fix late binding")
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: NCong Wang <xiyou.wangcong@gmail.com>
Acked-by: NJamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Now prio_init() can return -ENOMEM, it also has to make sure
any allocated qdiscs are freed, since the caller (qdisc_create()) wont
call ->destroy() handler for us.

More generally, we want a transactional behavior for "tc qdisc
change ...", so prio_tune() should not make modifications if
any error is returned.

It means that we must validate parameters and allocate missing qdisc(s)
before taking root qdisc lock exactly once, to not leave the prio qdisc
in an intermediate state.

Fixes: cbdf4511 ("net_sched: prio: properly report out of memory errors")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Reported-by: NCong Wang <xiyou.wangcong@gmail.com>
Acked-by: NCong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Maciej Żenczykowski reported lockdep warning a spinlock
was not registered before being held in mlx4_cmd_wake_completions()

cmd.context_lock initialization is not at the right place.

1) mlx4_cmd_use_events() can be called multiple times.
   Calling spin_lock_init() on a live spinlock can lead
   to hangs.

2) mlx4_cmd_wake_completions() can be called while lock
   has not been initialized.
   Lockdep complains, and current logic is not race prone.

It seems better to move the initialization earlier in
mlx4_load_one()
Signed-off-by: NEric Dumazet <edumazet@google.com>
Reported-by: NMaciej Żenczykowski <maze@google.com>
Cc: Eugenia Emantayev <eugenia@mellanox.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Nicolas Dichtel says:

====================
ovs: fix rtnl notifications on interface deletion

There was no rtnl notifications for interfaces (gre, vxlan, geneve) created
by ovs. This problem is fixed by adjusting the creation path.

v1 -> v2:
 - add patch #1 and #4
 - rework error handling in patch #2
====================
Acked-by: NPravin B Shelar <pshelar@ovn.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The function geneve_dev_create_fb() (only used by ovs) never calls
rtnl_configure_link(). The consequence is that dev->rtnl_link_state is
never set to RTNL_LINK_INITIALIZED.
During the deletion phase, the function rollback_registered_many() sends
a RTM_DELLINK only if dev->rtnl_link_state is set to RTNL_LINK_INITIALIZED.

Fixes: e305ac6c ("geneve: Add support to collect tunnel metadata.")
CC: Pravin B Shelar <pshelar@nicira.com>
CC: Jesse Gross <jesse@nicira.com>
CC: Thomas Graf <tgraf@suug.ch>
Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The function gretap_fb_dev_create() (only used by ovs) never calls
rtnl_configure_link(). The consequence is that dev->rtnl_link_state is
never set to RTNL_LINK_INITIALIZED.
During the deletion phase, the function rollback_registered_many() sends
a RTM_DELLINK only if dev->rtnl_link_state is set to RTNL_LINK_INITIALIZED.

Fixes: b2acd1dc ("openvswitch: Use regular GRE net_device instead of vport")
CC: Thomas Graf <tgraf@suug.ch>
CC: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The function vxlan_dev_create() (only used by ovs) never calls
rtnl_configure_link(). The consequence is that dev->rtnl_link_state is
never set to RTNL_LINK_INITIALIZED.
During the deletion phase, the function rollback_registered_many() sends
a RTM_DELLINK only if dev->rtnl_link_state is set to RTNL_LINK_INITIALIZED.

Note that the function vxlan_dev_create() is moved after the rtnl stuff so
that vxlan_dellink() can be called in this function.

Fixes: dcc38c03 ("openvswitch: Re-add CONFIG_OPENVSWITCH_VXLAN")
CC: Thomas Graf <tgraf@suug.ch>
CC: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

After ipgre_newlink()/geneve_configure() call, the netdev is registered.

Fixes: 7e059158 ("vxlan, gre, geneve: Set a large MTU on ovs-created tunnel devices")
CC: David Wragg <david@weave.works>
Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There is a corner case in which udp packets belonging to a same
flow are hashed to different socket when hslot->count changes from 10
to 11:

1) When hslot->count <= 10, __udp_lib_lookup() searches udp_table->hash,
and always passes 'daddr' to udp_ehashfn().

2) When hslot->count > 10, __udp_lib_lookup() searches udp_table->hash2,
but may pass 'INADDR_ANY' to udp_ehashfn() if the sockets are bound to
INADDR_ANY instead of some specific addr.

That means when hslot->count changes from 10 to 11, the hash calculated by
udp_ehashfn() is also changed, and the udp packets belonging to a same
flow will be hashed to different socket.

This is easily reproduced:
1) Create 10 udp sockets and bind all of them to 0.0.0.0:40000.
2) From the same host send udp packets to 127.0.0.1:40000, record the
socket index which receives the packets.
3) Create 1 more udp socket and bind it to 0.0.0.0:44096. The number 44096
is 40000 + UDP_HASH_SIZE(4096), this makes the new socket put into the
same hslot as the aformentioned 10 sockets, and makes the hslot->count
change from 10 to 11.
4) From the same host send udp packets to 127.0.0.1:40000, and the socket
index which receives the packets will be different from the one received
in step 2.
This should not happen as the socket bound to 0.0.0.0:44096 should not
change the behavior of the sockets bound to 0.0.0.0:40000.

It's the same case for IPv6, and this patch also fixes that.
Signed-off-by: NSu, Xuemin <suxm@chinanetcenter.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When the qdisc is full, we drop a packet at the head of the queue,
queue the current skb and return NET_XMIT_CN

Now we track backlog on upper qdiscs, we need to call
qdisc_tree_reduce_backlog(), even if the qlen did not change.

Fixes: 2ccccf5f ("net_sched: update hierarchical backlog too")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: WANG Cong <xiyou.wangcong@gmail.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: NCong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 26c5f03b uses a new skb allocator to avoid the RFD overflow
issue.

But from debugging without datasheet, we found the error always
happen when the DMA RX address is set to 0x....fc0, which is very
likely to be a HW/silicon problem.

So one idea is instead of adding a new allocator, why not just
hitting the right target by avaiding the error-prone DMA address?

This patch will actually
* Remove the commit 26c5f03b
* Apply rx skb with 64 bytes longer space, and if the allocated skb
  has a 0x...fc0 address, it will use skb_resever(skb, 64) to
  advance the address, so that the RX overflow can be avoided.

In theory this method should also apply to atl1c driver, which
I can't find anyone who can help to test on real devices.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=70761Signed-off-by: NFeng Tang <feng.tang@intel.com>
Suggested-by: NEric Dumazet <edumazet@google.com>
Tested-by: NOle Lukoie <olelukoie@mail.ru>
Acked-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Reported-by: NCong Wang <xiyou.wangcong@gmail.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Tom Herbert <tom@herbertland.com>
Fixes: 4068579e ("net: Implmement RFC 6936 (zero RX csums for UDP/IPv6")
Signed-off-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Cc: Tom Herbert <tom@herbertland.com>
Fixes: 4068579e ("net: Implmement RFC 6936 (zero RX csums for UDP/IPv6")
Signed-off-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Cc: Florent Fourcot <florent.fourcot@enst-bretagne.fr>
Fixes: 1d13a96c ("ipv6: tcp: fix flowlabel value in ACK messages send from TIME_WAIT")
Signed-off-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

IPv6 ping socket error handler doesn't correctly convert the new 32 bit
mtu to host endianness before using.

Cc: Lorenzo Colitti <lorenzo@google.com>
Fixes: 6d0bfe22 ("net: ipv6: Add IPv6 support to the ping socket.")
Signed-off-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Acked-by: NLorenzo Colitti <lorenzo@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Configuring the PHY LED registers for the Marvell 88E1510 and others is
not possible, because regardless of the values in marvell,reg-init, it
is later overridden in m88e1121_config_aneg with a non-standard default.

This patch moves that default configuration to .config_init to allow
setting the LED configuration through marvell,reg-init in the device
tree, which should override said default if it exists.
Signed-off-by: NClemens Gruber <clemens.gruber@pqgruber.com>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There is no reason to destroy channels that are destroyed while
cpdma_ctlr destroy. In this case no need to remember how much
channels where created and destroy them by one, as cpdma_ctlr
destroys all of them.
Signed-off-by: NIvan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

At Qdisc creation or change time, prio_tune() creates missing
pfifo qdiscs but does not return an error code if one
qdisc could not be allocated.

Leaving a qdisc in non operational state without telling user
anything about this problem is not good.

Also, testing if we replace something different than noop_qdisc
a second time makes no sense so I removed useless code.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

>> net/ipv4/ipconfig.c:130:15: warning: 'ic_addrservaddr' defined but not used [-Wunused-variable]
    static __be32 ic_addrservaddr = NONE; /* IP Address of the IP addresses'server */
Reported-by: Nkbuild test robot <fengguang.wu@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 7f854420
("phy: Add API for {un}registering an mdio device to a bus.")
broke PHY detection on this driver with a copy-paste bug:
The code is looking 32 times for a PHY at address 0.

Fixes ethernet on AMD DB1100/DB1500/DB1550 boards which have
their (autodetected) PHYs at address 31.

Cc: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: NManuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This reverts commit a2f27217.

I applied the wrong version of this.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 7f854420
("phy: Add API for {un}registering an mdio device to a bus.")
broke PHY detection on this driver with a copy-paste bug:
The code is looking 32 times for a PHY at address 0.

Fixes ethernet on AMD DB1100/DB1500/DB1550 boards which have
their (autodetected) PHYs at address 31.

Cc: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: NManuel Lauss <manuel.lauss@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

John Crispin says:

====================
net: mediatek: various small fixes

This series contains various small fixes that we stumbled across while
doing thorough testing and code level reviewing of the driver.

Changes in V2:
* drop the DQL patch from the list until a better solution is found
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The code checks if the queue should be stopped because we are below the
threshold of free descriptors only to check if it should be started again.
If we do end up in a state where we are at the threshold limit, it makes
more sense to just stop the queue and wait for the next IRQ to trigger the
TX housekeeping again. There is no rush in enqueuing the next packet, it
needs to wait for all the others in the queue to be dispatched first
anyway.
Signed-off-by: NJohn Crispin <john@phrozen.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The current code unconditionally wakes up the queue at the end of each
tx_poll action. Change the code to only wake up the queues if any of
them have actually been stopped before.
Signed-off-by: NJohn Crispin <john@phrozen.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The TX ring setup has an off by one error causing it to not utilise all
descriptors. This has the side effect that we need to reset the next
pointer at runtime to make it work. Fix the off by one and remove the
code fixing the ring at runtime.
Signed-off-by: NJohn Crispin <john@phrozen.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

During stress testing, after reducing the threshold value, we have seen
TX timeouts that were caused by the watchdog_timeo value being too low.
Increase the value to 5 * HZ which is a value commonly used by many other
drivers.
Signed-off-by: NJohn Crispin <john@phrozen.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The logic to calculate the threshold value for stopping the TX queue is
bad. Currently it will always use 1/2 of the rings size, which is way too
much. Set the threshold to MAX_SKB_FRAGS. This makes sure that the queue
is stopped when there is not enough room to accept an additional segment. 
Signed-off-by: NJohn Crispin <john@phrozen.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>